Re: [sched] fix sched_domains hotplug bootstrap ordering vs. cpu_online_map issue

[Nick Piggin <nickpiggin@yahoo.com.au>]

James Bottomley wrote:

>
>Well this patch got in, which is what I want, since it allows the
>non-NUMA machines to work with hotplug CPUs again.  However, is anyone
>actually looking to fix this for real?
>
>

I think someone else (tm) is looking at it :)
Some of the IBM hotplug guys I think.

>The fundamental problem is that NUMA or the scheduler (or both) are
>broken with regard to hotplug.
>
>The origin of the breakage is the differences between cpu_possible_map
>and cpu_online_map.  In hotplug CPU, there are two ways to do
>initialisations: you can initialise from cpu_online_map, but then you
>*must* have a cpu hotplug notify listener to add data structures for the
>extra CPUs as they come on-line, or you can initialise from
>cpu_possible_map and not bother with a notifier.  The disadvantage of
>the latter is that cpu_possible_map may be vastly larger than
>cpu_online_map ever gets to, thus wasting valuable kernel memory.
>
>The scheduler code is schizophrenic in this regard in that it does both:
>it initialises static data structures from cpu_possible_map, but it also
>has a hotplug cpu listener for starting things like the migration
>threads.
>
>I suspect the NUMA people would like us all to go to the former method
>(initialise only from cpu_online_map and have a proper hotplug listener)
>since their possible maps are pretty huge.  However, which is it to be:
>fix NUMA (to have two cpu_to_node() maps for the possible and online
>cpus per node) or fix the scheduler to do initialisation correctly?
>
>Perhaps this should be phased: change NUMA first temporarily for phase
>one and then fix the scheduler (and everyone else initialising from
>cpu_possible_map) in the second.
>
>

The scheduler *should* be able to be fixed nicely by using cpu_online_map
everywhere, and basically undoing then redoing the domains setup before and
after the hoplug, respectively.

So you'd re-attach the dummy domain to all CPUs, do the hotplug operation,
then setup the domains again and re-attach them.

This whole sequence could be pretty expensive, but I don't think the hotplug
guys care. It would allow us to get rid of cpus_and(... cpu_online_map) 
from a
lot of places in the scheduler too, which would be nice.

The actual code to do it shouldn't be more than a few lines (but I could be
overlooking something).